Three problems plague current methods of compound discovery. First, rational compound design can be used therapeutically to modify existing compounds if the compound target is known. However, for most diseases, there is little in depth understanding of the underlying biology. Second, compounds have also been selected based on their ability to inhibit or reverse specific disease-related phenotypes (e.g. protein proteolysis, aggregation). However, pre-selecting these phenotypes for compound development can be guesswork, as observable phenotypes do not often represent the primary defect or may be consequences of disease. Third, combinatorial library screens do not depend on knowing the compound target. However, these traditional screens are empirical in nature, and screening takes years, often without success. Even if a compound lead is identified in traditional screens, the actual compound target is typically not known. Therefore, increasing the potency and specificity of a lead in subsequent iterations can be a difficult and long-term process, and has not historically been successful. In this proposal, we develop, test and apply a new "intelligent" methodology that allows discovery in a systematic and directed manner. The method is called fingerprinting. The method is general and can apply to any disease gene, but in this proposal, we apply the methodology to Huntington's Disease. Disease-expressing cells are incubated with 20,000 siRNAs to knock-down all mammalian genes. Knocked down genes fall into two classes. Those that have no effect on toxicity and those genes whose loss enhance survival of mhtt-expressing cells. The set of genes whose loss enhances cell survival is called the gene fingerprint and defines genetic pathways associated with toxicity. The fingerprint siRNAs are inhibitors of toxicity since they remove gene pathways relevant to toxicity. If compounds also act as inhibitors of mhtt toxicity, they should act as an siRNA. Thus, the gene fingerprint defined by the siRNA should overlap with the fingerprint of a "good" chemical inhibitor. The genetic fingerprint can be "translated" into protein interactions to predict the actual targets of the inhibitor activity, and the pathways and targets are tested in models for disease. As an internal control, the fingerprinting methodology (Aim 2) is tested together with conventional screens to select for inhibitors to known disease- causing proteins (Aim 1). Inhibitors selected by the conventional methods are predicted to overlap with those identified from the global screen. [unreadable] [unreadable] [unreadable]